Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. An image reconstruction method comprising: obtaining image data, at least a portion of the image data relating to a region of interest (ROI); determining local information of the image data, wherein the local information including orientation information of the image data and gradient information of the image data; determining a regularization item based on a product of the orientation information of the image data and the gradient information of the image data, wherein the orientation information of the image data is modified by an Eigenvalue adjustment function that includes a factor of a scale of the Eigenvalues and at least one factor of a location of a peak of a characteristic curve; modifying the image data based on the regularization item; and generating an image based on the modified image data.
This invention relates to image reconstruction techniques, particularly for enhancing image quality in regions of interest (ROI) by incorporating local structural information. The method addresses challenges in image reconstruction where conventional approaches may produce artifacts or lose fine details, especially in medical imaging or other applications requiring high-resolution visualization. The process begins by acquiring image data, where at least part of the data corresponds to a specific ROI. Local information is then extracted from the image data, including orientation and gradient information, which describe the directional and intensity variations within the image. A regularization term is computed by combining the orientation and gradient information, with the orientation data being adjusted using an eigenvalue-based function. This function modifies the eigenvalues of the orientation information, incorporating both their scale and the position of a peak in a characteristic curve to refine the structural representation. The image data is subsequently adjusted using the computed regularization term, which helps preserve or enhance structural details while suppressing noise or artifacts. Finally, the modified image data is used to generate a reconstructed image with improved clarity and fidelity in the ROI. The technique is particularly useful in applications where maintaining fine structural details is critical, such as in medical imaging or high-precision industrial inspections.
2. The method of claim 1 , the determining a regularization item based on the local information comprising: smoothing the image data; determining orientation information of the smoothed image data; and determining the regularization item based on the gradient information of the image data and the orientation information of the smoothed image data.
This invention relates to image processing, specifically methods for determining a regularization item in image analysis or reconstruction tasks. The problem addressed is improving image quality by incorporating local image information to guide regularization, which helps reduce noise and artifacts while preserving important features. The method involves processing image data by first smoothing the image to reduce noise and enhance structural information. Orientation information is then extracted from the smoothed image, which describes the local directional features (e.g., edges or textures). A regularization item is computed by combining gradient information from the original image data with the orientation information derived from the smoothed image. This hybrid approach ensures that regularization adapts to the local image structure, avoiding over-smoothing of edges or loss of fine details. The technique is particularly useful in applications like medical imaging, where preserving anatomical structures is critical, or in computational imaging systems where raw data may be noisy. By dynamically adjusting regularization based on local image characteristics, the method improves reconstruction accuracy and visual quality. The smoothing step ensures robustness to noise, while the orientation-aware regularization preserves structural integrity. This approach can be integrated into iterative reconstruction algorithms or denoising pipelines to enhance performance.
3. The method of claim 2 , the smoothing the image data comprising smoothing the image data based on a first low-pass filter.
This invention relates to image processing techniques for enhancing image quality, particularly in systems where image data may contain noise or artifacts. The method involves smoothing image data to reduce noise while preserving important features. The smoothing process is performed using a first low-pass filter, which selectively attenuates high-frequency components associated with noise while retaining lower-frequency components that correspond to the desired image details. The low-pass filter may be applied in one or more dimensions, such as spatially across an image or temporally across a sequence of images. The method ensures that the smoothing operation does not excessively blur the image, maintaining clarity and sharpness in the processed output. This technique is particularly useful in applications where image quality is critical, such as medical imaging, surveillance, or high-resolution photography. The invention may also include additional preprocessing or postprocessing steps to further refine the image data before or after the smoothing operation. The overall goal is to improve image clarity and reduce visual artifacts without introducing distortion or loss of detail.
4. The method of claim 2 , the orientation information comprising a structure tensor of the smoothed image data or a modified structure tensor of the smoothed image data.
This invention relates to image processing techniques for analyzing and enhancing image data, particularly in the context of computer vision and machine learning applications. The problem addressed involves accurately determining orientation information from image data, which is crucial for tasks such as feature detection, image alignment, and texture analysis. Traditional methods often struggle with noise and inconsistencies in raw image data, leading to unreliable orientation estimates. The invention improves upon prior art by incorporating a structure tensor or a modified structure tensor derived from smoothed image data to compute orientation information. The structure tensor is a mathematical representation that captures local gradient information, helping to identify dominant directions in the image. By applying smoothing techniques to the image data before computing the structure tensor, the method reduces noise and enhances the accuracy of orientation estimation. The modified structure tensor may include additional processing steps, such as normalization or weighting, to further refine the orientation information. This approach is particularly useful in applications where precise orientation data is required, such as in medical imaging, autonomous navigation, and industrial inspection. The use of smoothed data ensures robustness against artifacts, while the structure tensor provides a computationally efficient way to extract meaningful orientation features. The invention can be integrated into existing image processing pipelines to improve the reliability of downstream tasks that depend on accurate orientation analysis.
5. The method of claim 4 , the determining orientation information of the smoothed image data comprising: determining the structure tensor of the smoothed image data; smoothing the structure tensor; determining Eigenvalues of the smoothed structure tensor; modifying the Eigenvalues; and determining a modified structure tensor based on the modified Eigenvalues.
This invention relates to image processing techniques for determining orientation information in image data, particularly for applications in computer vision, medical imaging, or autonomous systems where accurate feature detection is critical. The problem addressed is the challenge of reliably extracting orientation information from noisy or low-contrast image data, which can lead to inaccurate feature analysis. The method involves processing image data to enhance orientation detection by first smoothing the input image data to reduce noise. The smoothed image data is then used to compute a structure tensor, a mathematical representation that captures local gradient information and directional structures within the image. The structure tensor is further smoothed to refine the gradient information and reduce artifacts. Eigenvalues of the smoothed structure tensor are then calculated, which provide quantitative measures of the dominant orientation and strength of image features. These Eigenvalues are modified to emphasize or suppress certain orientations, depending on the application requirements. Finally, a modified structure tensor is reconstructed using the adjusted Eigenvalues, which improves the accuracy of orientation estimation. This approach enhances the robustness of orientation detection by systematically refining gradient information through tensor operations and eigenvalue adjustments, making it suitable for applications requiring precise feature analysis in challenging imaging conditions.
6. The method of claim 5 , the modifying the Eigenvalues comprising: determining the Eigenvalue adjustment function based on the at least a portion of the image data relating to the ROI; and revising the Eigenvalues based on the Eigenvalue adjustment function.
This invention relates to image processing, specifically techniques for modifying eigenvalues in image data to enhance or analyze regions of interest (ROIs). The problem addressed involves improving the accuracy or interpretability of image analysis by adjusting eigenvalues derived from image data, particularly in regions where specific features or anomalies are present. The method involves determining an eigenvalue adjustment function based on at least a portion of the image data corresponding to the ROI. This adjustment function is then applied to revise the eigenvalues, which are numerical values representing the variance or significance of data in the image. By modifying these eigenvalues, the method aims to refine the representation of the ROI, potentially improving tasks such as feature extraction, noise reduction, or anomaly detection. The eigenvalue adjustment function is derived from the image data itself, ensuring that the modifications are contextually relevant to the ROI. This approach allows for dynamic adaptation of the eigenvalues based on the specific characteristics of the image region being analyzed. The revised eigenvalues can then be used in subsequent image processing steps, such as dimensionality reduction, classification, or visualization, to enhance the overall analysis. This technique is particularly useful in applications where precise eigenvalue representation is critical, such as medical imaging, remote sensing, or quality control in manufacturing. By tailoring the eigenvalue adjustments to the ROI, the method provides a more accurate and efficient way to process and interpret image data.
7. The method of claim 6 , the ROI comprising a region relating to a liver, a bone, or a kidney.
This invention relates to medical imaging and image analysis, specifically for identifying regions of interest (ROIs) in medical images to improve diagnostic accuracy. The method involves analyzing medical images to detect and segment specific anatomical regions, such as the liver, bones, or kidneys, which are critical for diagnosing diseases like tumors, fractures, or organ dysfunction. The process includes preprocessing the image to enhance relevant features, applying segmentation algorithms to isolate the ROI, and validating the results to ensure accuracy. By focusing on these key anatomical regions, the method helps clinicians detect abnormalities more efficiently and reduce diagnostic errors. The technique is particularly useful in radiology, where precise identification of organs and structures is essential for treatment planning and monitoring. The invention improves upon existing methods by providing more reliable and automated segmentation, reducing the need for manual intervention and speeding up the diagnostic process.
8. The method of claim 5 , the determining the structure tensor of the smoothed image data comprising: determining a first-order differentiation of the smoothed image data; determining a transpose of the first-order differentiation of the smoothed image data; and determining the structure tensor of the smoothed image data based on the first-order differentiation of the smoothed image data and the transpose of the first-order differentiation of the smoothed image data.
This invention relates to image processing techniques for analyzing image data, specifically focusing on the computation of a structure tensor from smoothed image data. The structure tensor is a mathematical representation used in computer vision and image analysis to capture local image structures, such as edges, corners, and textures, which are essential for tasks like feature detection, image segmentation, and optical flow estimation. The method involves processing smoothed image data to determine its structure tensor. First, a first-order differentiation of the smoothed image data is computed, which involves calculating the gradient of the image intensity values. This gradient provides information about the rate of change in intensity, highlighting edges and transitions within the image. Next, the transpose of this first-order differentiation is determined, which rearranges the gradient components to facilitate further mathematical operations. The structure tensor is then derived by combining the first-order differentiation and its transpose, typically through matrix multiplication or other linear algebra operations. This tensor encapsulates the local gradient information in a structured form, enabling subsequent analysis of image features. The approach ensures that the structure tensor accurately represents the underlying image structures by leveraging smoothed data, which reduces noise and enhances feature clarity. This method is particularly useful in applications requiring robust feature extraction and analysis in noisy or complex image environments.
9. The method of claim 5 , the smoothing the structure tensor comprising applying a second low-pass filter on the structure tensor.
This invention relates to image processing techniques for smoothing structure tensors, which are used in computer vision and image analysis to represent local image structure. The problem addressed is the presence of noise or artifacts in structure tensors, which can degrade the accuracy of subsequent image processing tasks such as edge detection, texture analysis, or optical flow estimation. The method involves smoothing a structure tensor, which is a matrix representation derived from image gradients that captures local orientation and coherence information. The smoothing process is performed by applying a second low-pass filter to the structure tensor. Low-pass filtering removes high-frequency noise while preserving the underlying structure, improving the reliability of the tensor for further analysis. The first low-pass filter is applied to the image gradients before computing the structure tensor, while the second low-pass filter is applied directly to the structure tensor itself. This dual-filtering approach ensures that noise is effectively suppressed at both the gradient and tensor levels, enhancing the robustness of the image processing pipeline. The method is particularly useful in applications where high precision in structure estimation is critical, such as medical imaging, autonomous navigation, or industrial inspection.
10. The method of claim 2 , the determining gradient information of the image data comprising determining the gradient information based on a first-order differentiation of the image data.
This invention relates to image processing, specifically methods for analyzing gradient information in image data. The problem addressed is the need for accurate and efficient gradient computation, which is essential for tasks like edge detection, feature extraction, and image enhancement. Traditional gradient calculation methods often suffer from noise sensitivity or computational inefficiency, particularly in high-resolution or complex images. The method involves determining gradient information by performing a first-order differentiation of the image data. First-order differentiation is a mathematical technique that computes the rate of change of pixel intensity values, revealing edges and transitions within the image. This approach provides a computationally efficient way to extract gradient information, which can be used for further image analysis or processing tasks. The method may be applied to various types of image data, including grayscale, color, or multi-spectral images, and can be implemented in hardware or software systems. By using first-order differentiation, the method ensures that gradient information is derived directly from the raw image data, minimizing errors introduced by higher-order approximations. This technique is particularly useful in real-time applications where processing speed is critical, such as medical imaging, autonomous navigation, or industrial inspection. The method can be combined with other image processing techniques, such as filtering or segmentation, to enhance overall system performance.
11. The method of claim 1 , the obtaining image data comprising: obtaining projection data; and generating the image data based on the projection data.
This invention relates to medical imaging, specifically methods for generating image data from projection data. The problem addressed is the need for accurate and efficient image reconstruction in imaging systems such as computed tomography (CT) or magnetic resonance imaging (MRI), where raw projection data must be processed into usable images. The method involves obtaining projection data, which consists of raw measurements captured by imaging sensors. These measurements represent the attenuation or signal response of a scanned object from multiple angles. The projection data is then processed to generate image data, which reconstructs a detailed representation of the scanned object. This reconstruction may involve mathematical algorithms such as filtered back projection or iterative reconstruction techniques to convert the projection data into a high-resolution image. The method ensures that the image data accurately reflects the internal structure of the scanned object, improving diagnostic accuracy in medical imaging. The approach is particularly useful in applications requiring precise visualization of anatomical features or pathological conditions. By optimizing the conversion of projection data into image data, the method enhances the reliability and clarity of medical imaging results.
12. The method of claim 11 , the generating the image data based on the projection data comprising updating the image data based on an iterative statistical reconstruction algorithm.
This invention relates to medical imaging, specifically to methods for generating image data from projection data, such as in computed tomography (CT) or other tomographic imaging techniques. The problem addressed is improving the accuracy and quality of reconstructed images by refining the image data through iterative statistical reconstruction algorithms. These algorithms iteratively adjust the image data to better match the measured projection data, reducing artifacts and noise while preserving fine details. The method involves acquiring projection data from a scanning system, where the projection data represents measurements of radiation or other signals passing through an object from multiple angles. The initial image data is generated from this projection data, typically using a reconstruction technique such as filtered back-projection or another initial reconstruction method. The key improvement lies in updating this initial image data through an iterative statistical reconstruction algorithm. This algorithm statistically models the noise and system characteristics, iteratively refining the image data to minimize discrepancies between the reconstructed image and the measured projection data. The iterative process may incorporate prior knowledge, such as anatomical constraints or statistical models of image features, to enhance reconstruction accuracy. By using this approach, the method improves image quality, particularly in low-dose or sparse-data scenarios, where traditional reconstruction methods may produce poor results. The iterative statistical reconstruction helps suppress noise and artifacts while preserving diagnostic details, making it valuable for medical imaging applications where high-quality images are critical.
13. The method of claim 1 , the image data comprising a 2D image, 2D image data, a 3D image, or 3D image data.
This invention relates to image processing techniques for analyzing and interpreting image data, which may include two-dimensional (2D) images, 2D image data, three-dimensional (3D) images, or 3D image data. The method addresses the challenge of accurately processing and extracting meaningful information from various types of image data, which can be complex due to differences in dimensionality, resolution, and structural characteristics. The core technique involves a computational process that adapts to the specific type of image data being analyzed. For 2D images, the method may involve pixel-level analysis, feature extraction, or pattern recognition to identify objects, edges, or other relevant features. For 3D images, the method extends this analysis to volumetric data, incorporating depth information to reconstruct or interpret three-dimensional structures. The approach may also include preprocessing steps to enhance image quality, such as noise reduction, contrast adjustment, or normalization, ensuring consistent and reliable results across different image types. Additionally, the method may integrate machine learning or artificial intelligence techniques to improve accuracy and efficiency in image interpretation. By leveraging these advanced algorithms, the system can adapt to diverse imaging scenarios, making it suitable for applications in medical imaging, industrial inspection, autonomous navigation, or other fields where precise image analysis is critical. The flexibility to handle both 2D and 3D data ensures broad applicability across various domains.
14. An image reconstruction method comprising: obtaining image data, at least a portion of the image data relating to a region of interest (ROI); determining gradient information of the image data; determining a structure tensor of the image data; determining a regularization item based on a product of the gradient information and the structure tensor, wherein the structure tensor is modified by an Eigenvalue adjustment function that includes a factor of a scale of the Eigenvalues and at least one factor of a location of a peak of a characteristic curve; modifying the image data based on the regularization item; and generating an image based on the modified image data.
This invention relates to image reconstruction techniques, specifically addressing challenges in enhancing image quality by reducing noise and preserving structural details. The method involves processing image data, where at least part of the data corresponds to a region of interest (ROI). Gradient information of the image data is computed to analyze local variations, while a structure tensor is derived to capture directional structures within the image. A regularization term is then calculated by combining the gradient information with the structure tensor, which is adjusted using an eigenvalue modification function. This function incorporates a scaling factor for the eigenvalues and a factor based on the peak location of a characteristic curve, allowing fine-tuned control over noise suppression and feature preservation. The image data is subsequently modified using this regularization term, and a final image is generated from the processed data. The technique aims to improve image clarity by balancing noise reduction with the retention of important structural features, particularly in regions of interest.
15. The method of claim 14 , the determining a structure tensor of the image data comprising: optimizing the structure tensor by smoothing algorithm or/and modifying algorithm.
Technical Summary: This invention relates to image processing, specifically methods for analyzing image data to determine structural features. The core problem addressed is the accurate extraction of structural information from images, which is essential for applications like computer vision, medical imaging, and autonomous systems. The invention focuses on improving the computation of a structure tensor, a mathematical representation used to analyze local image structures such as edges, corners, and textures. The method involves optimizing the structure tensor through two key approaches: smoothing algorithms and modifying algorithms. Smoothing algorithms reduce noise and enhance structural clarity by applying filters or other techniques to refine the tensor. Modifying algorithms adjust the tensor to better represent the underlying image features, potentially by incorporating additional constraints or adaptive parameters. These optimizations ensure that the structure tensor more accurately reflects the true structural characteristics of the image, improving downstream tasks like feature detection, segmentation, and tracking. The invention builds on prior techniques by introducing flexibility in how the structure tensor is computed, allowing for tailored optimizations based on the specific requirements of the application. This adaptability enhances the robustness and accuracy of image analysis in diverse scenarios.
16. The method of claim 15 , wherein the smoothing algorithm includes Gaussian filter, and the modifying algorithm includes the Eigenvalue adjustment function.
This invention relates to image processing techniques for enhancing image quality, particularly in medical imaging applications. The method addresses the challenge of noise reduction and feature preservation in images, where traditional filtering techniques often either fail to adequately remove noise or inadvertently distort important image features. The invention provides a solution by combining a smoothing algorithm with a modifying algorithm to improve image clarity while maintaining critical structural details. The smoothing algorithm applies a Gaussian filter to reduce noise in the image. Gaussian filters are effective for blurring and smoothing by convolving the image with a Gaussian function, which attenuates high-frequency noise while preserving low-frequency image content. However, excessive smoothing can blur edges and fine details, which is problematic in medical imaging where precise feature detection is essential. To counteract this, the modifying algorithm incorporates an Eigenvalue adjustment function. This function selectively adjusts the eigenvalues of the image data, which represent the variance in different directions. By modifying these eigenvalues, the algorithm enhances contrast and sharpness in key regions while suppressing noise. The Eigenvalue adjustment function ensures that important structural features remain distinct, even after smoothing. The combined approach of Gaussian filtering followed by Eigenvalue adjustment provides a balanced solution, improving image quality by reducing noise while preserving critical details. This method is particularly useful in medical imaging, where accurate diagnosis relies on clear and detailed images. The technique can be applied to various imaging modalities, including MRI, CT scans, and ultrasound, to
17. A system, comprising: at least one storage medium including a set of instructions for image reconstruction; and at least one processor configured to communicate with the at least one storage medium, wherein the set of instructions, when executed by the at least one processor, cause the system to perform operations including: obtaining image data, at least a portion of the image data relating to a region of interest (ROI); determining local information of the image data, the local information including orientation information of the image data and gradient information of the image data; determining a regularization item based on a product of the orientation information of the image data and the gradient information of the image data, wherein the orientation information of the image data is modified by an Eigenvalue adjustment function that includes a factor of a scale of the Eigenvalues and at least one factor of a location of a peak of a characteristic curve; modifying the image data based on the regularization item; and generating an image based on the modified image data.
The system is designed for image reconstruction, particularly for enhancing image quality in regions of interest (ROI) by applying adaptive regularization techniques. The problem addressed is the need for improved image reconstruction methods that preserve fine details while reducing noise and artifacts, especially in medical imaging or other applications where image clarity is critical. The system includes a storage medium containing instructions and at least one processor to execute those instructions. The process begins by obtaining image data, where at least part of the data corresponds to a specific ROI. The system then analyzes the image data to extract local information, including orientation and gradient information, which describe the structural and intensity variations within the image. A key aspect of the system is the determination of a regularization item, which is derived from the product of the orientation and gradient information. The orientation information is adjusted using an eigenvalue adjustment function that incorporates both the scale of the eigenvalues and the location of a peak in a characteristic curve. This adjustment helps tailor the regularization to the local image features, ensuring that the reconstruction process adapts to the underlying image structure. The image data is then modified based on this regularization item, and a final image is generated from the modified data. This approach aims to improve image quality by selectively enhancing relevant features while suppressing noise, particularly in regions of interest. The adaptive nature of the regularization ensures that the reconstruction process is optimized for the specific characteristics of the input image data.
18. The system of claim 17 , the operations further including: smoothing the image data; determining orientation information of the smoothed image data; and determining the regularization item based on the gradient information of the image data and the orientation information of the smoothed image data.
This invention relates to image processing systems designed to enhance image quality by reducing noise and artifacts while preserving structural details. The system processes image data by first smoothing the data to reduce noise. After smoothing, the system determines orientation information, which identifies dominant edge directions in the image. The system then calculates a regularization item—a mathematical term used to refine the image—based on both the gradient information of the original image data and the orientation information derived from the smoothed data. The gradient information captures fine details and edges, while the orientation information ensures that smoothing is applied in a direction-aligned manner, preventing over-blurring of important features. This approach improves image clarity by balancing noise reduction with detail preservation, particularly useful in medical imaging, satellite imagery, or other applications where both noise suppression and structural accuracy are critical. The system may be part of a larger image reconstruction or enhancement pipeline, where the regularization item is integrated into an optimization process to refine the final output. The method ensures that smoothing operations adapt to the local structure of the image, avoiding distortion of edges and textures.
19. The system of claim 18 , the operations further including: determining a structure tensor of the smoothed image data; smoothing the structure tensor; determining Eigenvalues of the smoothed structure tensor; modifying the Eigenvalues; and determining a modified structure tensor based on the modified Eigenvalues.
This invention relates to image processing, specifically to enhancing image features by analyzing and modifying the structure tensor of image data. The system processes an input image to improve feature detection and representation, particularly in applications like edge preservation, texture analysis, or noise reduction. The system first smooths the input image data to reduce noise while preserving important features. It then computes a structure tensor from the smoothed image data, which captures local image gradients and orientation information. The structure tensor is further smoothed to enhance robustness against noise. Eigenvalues of the smoothed structure tensor are then calculated, representing the strength and directionality of image features. These Eigenvalues are modified to emphasize or suppress certain features, such as enhancing edges or reducing noise. Finally, a modified structure tensor is reconstructed using the adjusted Eigenvalues, which can be used for further image analysis or processing tasks. This approach improves feature extraction by dynamically adjusting the structure tensor's properties, leading to better performance in tasks requiring accurate gradient and orientation information. The method is particularly useful in applications where preserving fine details or distinguishing between noise and meaningful features is critical.
20. The system of claim 17 , the operations further including: determining the Eigenvalue adjustment function based on the at least a portion of the image data relating to the ROI; and revising the Eigenvalues based on the Eigenvalue adjustment function.
This invention relates to image processing systems that analyze regions of interest (ROIs) within images. The system addresses the challenge of accurately processing and interpreting image data, particularly when dealing with variations in ROI characteristics. The system includes a computational module that processes image data to identify and extract features from the ROI. It employs Eigenvalue-based analysis to represent these features mathematically, which is crucial for tasks like object recognition, medical imaging, or quality control in manufacturing. The system further includes operations to dynamically adjust the Eigenvalues used in the analysis. This adjustment is based on a portion of the image data specific to the ROI, ensuring that the mathematical representation remains accurate and relevant to the image content. The Eigenvalue adjustment function is determined by analyzing the ROI data, allowing the system to adapt to variations in image quality, lighting conditions, or structural differences within the ROI. The Eigenvalues are then revised according to this function, improving the system's ability to extract meaningful features and enhance processing accuracy. This adaptive approach ensures that the Eigenvalue-based analysis remains robust across different imaging scenarios, making the system more reliable for applications requiring precise feature extraction and interpretation. The dynamic adjustment of Eigenvalues based on ROI-specific data improves the system's flexibility and accuracy in real-world applications.
Unknown
February 11, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.